The original depth rendering

[cookie-yu]

Thank you for your amazing work! But I had a question, I noticed that you mentioned in your paper: "We also extend the original 3DGS to render depth" So which method do you use to render 3DGS depth? Do you render in this way?

[hbb1]

We indeed adopt the mean depth for 3DGS.

[cookie-yu]

Thank you so much! And when I try to extract the depth information of 3DGS and 2DGS, I found the performace of 2DGS is much better. Just like the following two images.


The second image depth display in flat areas is smoother. What do you think is the main reason for this phenomenon? Why the depth information of 3DGS is not correct?

[hbb1]

Ray-splat intersection scheme has better depth.